Session-by-session notes. What I tried. What worked. What didn’t. A breadboard reliability saga.
This is the technical companion to Ponderations — terse, dated, one entry per session. When a piece wants more than that — a thesis, an audit, an idea worth a thousand words — it lives over there instead.
To be written
Sessions worth writing up, in the order they happened:
- 2026-03-24 — Vision doc drafted. Basement state inventoried.
- 2026-03-27 — First real data. Atlas pH probe live at I2C 0x63, DS18B20 temp on GPIO4.
- 2026-04-07 — Flood tray installed. Would get reversed 12 days later.
- 2026-04-10 — Autonomous heater control confirmed working. Relay flipping based on temp thresholds.
- 2026-04-14 — Inventory + bug triage. Heater overshoot, timezone display, duplicate scheduler.
- 2026-04-19 — The DWC switch. Flood tray out, DWC buckets in. First real transplant.
- 2026-04-20 — Vision doc reconciled with current reality. JARVIS Phase 5 recognition.
Why a build log
Three reasons.
First — documentation that the work is actually happening. The difference between “we’re planning an aquaponics project” and “Node Zero has been logging data every 30 seconds since March 27” is the difference between a pitch deck and a product.
Second — the failures are the interesting part. Most aquaponics content online shows finished systems. The log is the unfinished ones. The flood tray that got installed and then pulled out. The duplicate scheduler that’s been corrupting half the pH readings since day one. The breadboard on the carpet that keeps dropping sensors.
Third — this is the raw material for the AI assistant. Every session written down is something Start can draw on when someone asks “has anyone tried X before?”